PI SQC User Guide
Contents
1. E 23 POLL aaa 25 Appendix A Statistical Calculation Basis 52226222 27 Control Chart Calculations aaa amaya a na d 27 Ree m 36 Mathematical 1006 OE Tm 39 Process Capability Index COOK sese eee 40 Appendix B GIOSSSEV a USA ORIS oo Ren ee Cn HI Dex NUR DCN ERR NOU eee OP CA RR E NERO 41 Appendix C Technical Support and Resources 88 8888888888888 43 a ean 47 PI SQC User Guide Chapter 1 Introduction to PI SQC PI SQC 1s a PI ProcessBook add in that retrieves data from PI archives and ODBC databases and performs statistical calculations to determine data behavior Results are displayed in a three part PI SQC chart which includes one of eight types of control charts The PI SQC chart appears within a standard PI ProcessBook display or independent display file Statistical Quality Control Statistical Quality Control SQC has at its heart the experience that all processes fluctuate Fluctuations may be natural or unnatural Natural fluctuations are generally small while unnatural fluctuations are larger and are introduced by external and hopefully identifiable causes In simple terms e Everything Varies People live to different ages all patterns fluctuate e I2. Calculation Period For Time based calculations the Calculation Period is the time interval between the start of sample groups The time period must be entered in the form hh mm ss where hh is hours mm is minutes and ss is seconds Calculation Period is ignored for Event based calculations Sample Period For Time based calculations the Sample Period is the time between samples within a sample group The product of the Sample Period and the Sample Size should be less than or equal to the Calculation Period For example if the Calculation Period is 2 hours the Sample Period is 10 minutes and the Sample Size is 5 the sample groups would begin at midnight 2 00 4 00 and so on The samples for the midnight sample group would be at 12 00 12 10 12 20 12 30 and 12 40 The samples for the next group would be at 2 00 2 10 2 20 2 30 and 2 40 Start Time of Sample Use Start Time of Sample to adjust the start time of a sample within a day If the Calculation Period is greater than one day the Start Time is used for the day on which the calculation is started The derived values are calculated and stored at the time of the last sample within the Calculation Period For example if the Calculation Period is 2 hours the Sample Period is 10 minutes the Sample Size is 5 and the Start Time is 3 a m then sample groups begin at 3 a m 5 a m and so forth Samples for the first sample group would be taken at 3 00 3 10
3. On the control chart the timestamp defining the left end of the horizontal scale represents the first value derived after the trigger tag value changes indicating the start of a product run The plotted point at the right end 1s either the last value before the current time for real time data or the last plotted value before the specified end time of the plot and product run os Sample Tab Note The time of a trigger tag change does not equate to the start time of a sample but instead determines which data point demarcates the first point of a sample group The start time of the actual sample is defined on the Sample tab and occurs after the trigger tag change If the time range or its display is changed the new end time is used to determine the time of the previous trigger tag change End Time is defined using standard PI time relative expressions measuring time in multiples of hours and days from the present time For charts that end at the current time whenever a PI tag value changes or a new measurement is received from an ODBC data set data is recalculated and both the control chart and histogram are updated Sample Tab The Sample tab defines how chart tag samples are calculated Sampling Method Two methods can be selected to work raw data into packets for calculating samples Event based sampling Event based sampling is depends on a sample size n where every n raw data points are combin
4. e One Side of Centerline A count of the number of data points on one side of the centerline Stratification A count of the number of data points that fall within the upper and lower one sigma limits on both sides of the center line e Mixture A count of the number of data points that fall outside the upper and lower one sigma limits on both sides of the center line e Trend A count of the number of data points which are monotonically increasing or decreasing To configure pattern tests e Click the checkbox at left to enable a test e Click the marker symbol in the next column and choose Marker to reset the marker shape or Color to reset the marker color through the Windows color palette The marker format appears against zone backgrounds set on the Format tab page 25 e Enter values to describe the test conditions in the two entry fields to complete the test test condition Alarm Reset Y ou may also specify an Alarm Reset tag by clicking the ellipsis button and using the Tag Search dialog to select a PI tag When the specified tag reaches zero alarm calculations are stopped When the Alarm Reset tag is no longer zero alarm calculations restart If no reset tag 1s specified alarms are always on by default Pattern Tests Alarm pattern tests compare derived values to control limit lines center line and intermediate values The simplest pattern is a test for a single value outside of control limits
5. Control charts are more sensitive to process changes when more complicated tests are used Such tests use a sequence of values to test for patterns that are unlikely in a normal os Format Tab distribution For example an out of control condition for charts may be when 2 out of 3 consecutive values are outside of 2 Sigma Default settings for pattern tests are based on recommendations from the Statistical Quality Control Handbook Western Electric Co Inc 1956 Indianapolis Indiana on pages 23 30 under Tests for Unnatural Patterns Format Tab The Format tab provides tools used to modify the appearance of SQC charts As you make changes in the Format tab the sample chart at the lower right of the tab changes to preview your selections Chart Element Select an element and use the accompanying options to specify Line Style and Width Data Markers and their Size and the Color of the selected element Center Zone 0 1 Sigma above or below the center line One Sigma Zone between 1 2 Sigma above or below the center line Two Sigma Zone between 2 3 Sigma above or below the center line Outside 3 Sigma Zone 3 Sigma above or below the center line Text on all parts of the chart The font size and font style are configurable using the PI ProcessBook font toolbar button Background color of the chart background Control Chart Data pen color and line style for the curve plotted on the control chart Specificat
6. osi PI SQC User Guide Version 3 2 OSIsoft Inc 777 Davis St Suite 250 San Leandro CA 94577 USA Additional Offices Houston TX Johnson City TN Longview TX Mayfield Heights OH Philadelphia PA Phoenix AZ Savannah GA Sales Outlets Distributors Middle East North Africa Republic of South Africa Russia Central Asia South America Caribbean Southeast Asia South Korea Taiwan Contact and Support Main phone Fax Support phone Web site Support web site Support email International Offices OSIsoft Australia Perth Australia Auckland New Zealand OSIsoft Germany GmbH Altenstadt Germany OSIsoft Asia Pte Ltd Singapore OSIsoft Canada ULC Montreal Canada Calgary Canada OSIsoft Inc Representative Office Shanghai People s Republic of China OSIsoft Japan KK Tokyo Japan OSIsoft Mexico S De R L De C V Mexico City Mexico OSIsoft do Brasil Sistemas Ltda Sao Paulo Brazil 01 510 297 5800 01 510 357 8136 01 510 297 5828 http www osisoft com http techsupport osisoft com techsupport osisoft com Copyright 1992 2009 OSlsoft Inc All rights reserved No part of this publication may be reproduced stored in a retrieval system or transmitted in any form or by any means mechanical photocopying recording or otherwise without the prior written permission of OSlsoft Inc OSIsoft the OSlsoft logo and logotype PI Analytics PI ProcessBook PI DataLink P
7. 3 20 3 30 and 3 40 Time Stamp at Specify the raw data point to use in determining a timestamp for each calculated point This feature is useful when the Sample Size is larger than 1 For example if the sample size is 5 and Time stamp at is set to 3 the time at which every third raw data point 1s collected 1s used as the timestamp for the related calculated point Lambda Lambda associates a weight with each new data point as compared with the remaining data points in the designated sample The Lambda factor pertains only to EWMA page 13 Exponentially Weighted Moving Average charts and does not appear for other types PI SQC User Guide 19 PI SQC Chart Definition Dialog 20 Lambda must be between 0 and 1 A smaller value of Lambda provides greater smoothing effects and the resulting chart will emphasize trends as opposed to point to point fluctuations The default value for Lambda is 0 5 with values between 0 2 and 0 5 most typical Filter Equations Filter Equations can be used to filter data from samples In a filter equation 0 equals FALSE and 1 equals TRUE Only data representative of time periods when the Filter Equation is TRUE are included in samples and used in PI SQC calculations Note Filter equations can include PI tags but not ODBC data sets However you can filter ODBC data using the WHERE clause of the ODBC query Use the following tools to build a filter equation in the Filter Equation fie
8. You may choose to keep bad values in control charts allowing PI tag values with invalid status to be included in sample groups but not in calculations In such cases the chart is calculated using only the remaining good values If bad values are kept they are counted when creating sample subgroups for averages ranges and so forth The number of markers is then the total number of values good and bad divided by the sample size with any remainder ignored If bad values are not kept but rather omitted from the time series of values all samples are the same size For example if Keep Bad Values is checked and if the sample size is 5 and if a group of 5 data points includes one bad point only 4 values will be used for the calculation for this particular plot point PI values that fail the criteria are plotted with a black dot on the Shewhart chart Ifa consecutive group of values fail the filter criteria only the first PI value to fail is plotted repeated consecutive failures are omitted from the sample An entire sample calculation 1s determined to be bad and 1s not plotted whenever the number of good values is less than 50 of the specified sample size The Range and Moving Range calculations are exceptions in that they require only 2 good values in a sample subgroup Note For EWMA charts a bad value resets the calculation If EVVMA is bad EWMA Xp Control Parameters Tab Control parameters are process limits w
9. Range R Chart A Range chart calculates the range maximum minimum within each sample to determine the stability of a process Calculated ranges are used to estimate process variability due to chance causes over short time spans For example a valve linkage in need of repair does not move the same way each time it 1s told to move This inconsistency in valve action affects the product differently at different times and might be examined through a Range chart A Range chart 1s always evaluated in conjunction with either an Individuals chart or an Average chart Sigma page 7 values may be interpreted differently for range charts Note Range charts are not recommended for sample sizes larger than 8 and require a sample size greater than 1 with a sample size of 2 as the default value Moving Range Rm Chart A Moving Range chart uses the latest n observations to judge the stability of the process The current range 1s calculated by replacing the oldest value in the sample with the current value As with Moving Average charts the effect of combining successive data points causes correlated contributions of random error and results in an oscillatory effect when plotted Sigma page 7 values may also be interpreted differently for range charts Note Moving Range charts are not recommended for sample sizes larger than 8 Chart Tag Enter the name of the source tag to plot in the chart The Chart Tag specifies the proces
10. create a new ODBC data set Click the arrow button and select Element Context to select an AF Element as a control limit Note this option only appears if you use PI SQC with PI ProcessBook 3 2 or higher Click one of the Sigma Calculation Method options o Average Range Estimate o Average Standard Deviation Estimate When a new SQC symbol 15 created through the user interface or automation the default values of the Sigma Calculation Method vary according to the plot type The plot type also determines if the Sigma Calculation Method option buttons are disabled or non modifiable through automation These buttons are also disabled for server side alarms regardless of plot type You may also let PI SQC calculate limits based on the data in the plot If you do not specify a Center Line value PI SQC uses the mean of the plot data as the center line If you do not specify any control limit parameters PI SQC calculates limits based on a Sigma value determined by the distribution of data Data Filter Tab Data filters allow removal of data that do not represent a nominal process Data that do not meet the specified criteria are discarded and standard deviation process capability mean and all other statistical calculations are then made with the remaining data Statistical calculations are generally computed at the end of the defined time period governed by a trigger tag page 16 The defined time period is always the time p
11. performs a calculation for every Calculation Period determined whenever samples matching the Sample Size are collected e An Event based calculation uses every raw data value associated with the tag For event based calculations Calculation Period Sample Period and Start Time of Sample are disabled Note Calculation Basis information is not available for PISQC RT Alarm Tags Check Settings Click Check Settings to test time settings for Time based samples and ensure the minimum sample size once you have specified all the necessary parameters for your calculation basis Sample Size The Sample Size is the number of measurements grouped into one sample samples should be chosen such that each sample encompasses all the random variation inherent in the process A sampling scheme that meets this criterion is said to produce rational subgroups Provided samples are aligned with rational subgroups the smaller the sample size the more sensitive calculations are to process changes For new charts the minimum sample size is the default size The following minimum and recommended maximum values apply to sampling Chart Type Default Recommended Minimum Maximum Individual 1 X bar or Average 2 Moving Average 2 EWMA 1 Range 2 8 Moving Range 2 8 Standard Deviation 8 osi Sample Tab Chart Type Default Recommended Minimum Maximum Moving Std Deviation 8
12. process is a set of conditions or causes that work together to produce a result In an industrial setting a process can be a single control loop a unit operation a laboratory measurement a task performed by a single person or a team or virtually any combination of actors which work together to produce a result Sample Grouping Breaking a large collection of measurements into subgroups Evaluating sample averages and ranges provides a more sensitive tool for spotting process variations with definable causes unnatural patterns Tests for Unnatural Patterns There are two basic premises for testing the naturalness of a pattern o Test individual points versus limits o Test multiple points for trend there are several commonly accepted tests for trend recognition e g 8 points in a row on one side of the plot center line Unnatural patterns indicate possible process variations with definable causes PI SQC User Guide 41 Glossary 42 Unnatural Fluctuation These are large abnormal fluctuations due to some external cause The causes of unnatural fluctuations are generally definable to some outside influence For example an unnatural fluctuation could arise from software malfunction field instrument failure putting the wrong material into the product or similar causes Unnatural Pattern A pattern of measurements of a process parameter exhibiting large fluctuations due to a definable cause An unnatural pattern will
13. server name will be included PI SQC uses ODBC data sets and custom placeholders in the same manner as PI ProcessBook trends You can use an alias or property value or a dataset in place of a tag e Click the adjacent arrow button and choose a PI Calculation or ODBC Data Set from the menu to search other data sources using a corresponding Data dialog If an ODBC Data Set is selected then the SOL query must contain the time placeholders start Time and End Time in the ODBC dataset to return the time frame specified in the SQC chart definition If a dataset includes default placeholders you can click Custom Placeholders to supply values for variables used in the ODBC query e Choose Module Context to select aliases or properties via a module context defined in a PI Module Database Scale Use Scale group fields to set the minimum and maximum for the chart s two vertical scales Scales are based either on data values explicit limits or standard deviation calculations The center line or target is the mean value of the plot data PI SQC User Guide 15 PI SQC Chart Definition Dialog The following scale options are available AutoRange the default which automatically sets chart limits at even numbers near the minimum and the maximum of plotted values Database for PI tags only which uses the zero and span attributes of the PI SQC tag plotted Zero and span are parameters configured in PI and cannot be changed locally
14. the stability of the process An Individuals chart is used when it is impractical to calculate the average of a group of measurements Individuals charts are typically used in conjunction with Range or Moving Range charts to detect variability In Individuals charts sample size must be equal to 1 and the center line is the process mean The Moving Range is used to estimate standard deviation when calculating control limits If a PI tag 1s used as the upper USL or lower specification limit LSL the control chart displays USL and LSL traces Note Measurements are not independent of one another in an Individuals chart Pattern tests are therefore not applicable and only the points outside of control limits should be considered significant X Bar Average Chart An X Bar or Average chart calculates the average of a subgroup of data and plots it against other sample averages The sample mean is used as the center line unless otherwise specified Individual measurements are assumed to be independent An X Bar chart indicates where a process is centered and provides information about the quality of process control Fluctuations that directly affect the X Bar chart display can be construed to affect all parts of the product at once and in the same general manner Averaging data tends to reduce the effect of random error associated with the variability due to measurement and inherent process variation Averages can be mor
15. 0 313 1 637 12 0 9776 0 354 1 646 0 346 1 610 13 0 9794 0 382 1 618 0 374 1 585 14 0 9810 0 406 1 594 0 399 1 563 15 0 9823 0 428 1 572 0 421 1 544 16 0 9835 0 448 1 552 0 440 1 526 os Mathematical Symbols c4 B3 B4 B5 B6 17 0 9845 0 466 1 534 0 458 1 511 18 0 9854 0 482 1 518 0 475 1 496 19 0 9862 0 497 1 503 0 490 1 483 20 0 9869 0 510 1 490 0 504 1 470 21 0 9876 0 523 1 477 0 516 1 459 22 0 9882 0 534 1 466 0 528 1 448 23 0 9887 0 545 1 455 0 539 1 438 24 0 9892 0 555 1 445 0 549 1 429 25 0 9896 0 565 1 435 0 559 1 420 Mathematical Symbols The following mathematical symbols are used in calculations M Number of Samples n Number of measurements in a Sample g Number of samples Sample Range K Average Range Average Moving Range Po sample standard deviation 5 Average sample standard deviation Xmax Maximum measurement within a sample Xmin Minimum measurement within a sample Individual Observation Xi Sample Average i l Average of Samples R i l rol Grand Average of Samples PI SQC User Guide 39 Statistical Calculation Basis 40 Weighting Factor for EWMA charts Process Capability Index Cpk The Process Capabili
16. 534 0 5 078 0 2 004 7 2 104 0 204 5 204 0 076 1 924 8 2 847 0 388 5 306 0 136 1 864 9 2 970 0 547 5 393 0 184 1 816 10 3 078 0 687 5 469 0 223 1 777 PI SQC User Guide Factors 37 Statistical Calculation Basis 38 n d2 D1 D2 D3 D4 11 3 173 0 811 5 535 0 256 1 744 12 3 258 0 922 5 594 0 283 1 717 13 3 336 1 025 5 647 0 307 1 693 14 3 407 1 118 5 696 0 328 1 672 15 3 472 1 203 5 741 0 347 1 653 16 3 532 1 282 5 782 0 363 1 637 17 3 588 1 356 5 820 0 378 1 622 18 3 640 1 424 5 856 0 391 1 608 19 3 689 1 487 5 891 0 403 1 597 20 3 735 1 549 5 921 0 415 1 585 21 3 778 1 605 5 951 0 425 1 575 22 3 819 1 659 5 979 0 434 1 566 23 3 858 1 710 6 006 0 443 1 557 24 3 895 1 759 6 031 0 451 1 548 25 3 931 1 806 6 056 0 459 1 541 Factors for Standard Deviations Observations in Sample n n c4 B3 B4 B5 B6 2 0 7979 O 3 267 2 606 3 0 8862 0 2 568 0 2 276 4 0 9123 0 2 266 0 2 088 5 0 9400 0 2 089 0 1 964 6 0 9515 0 030 1 970 0 029 1 874 7 0 9594 0 118 1 882 0 113 1 806 8 0 9650 0 185 1 815 0 176 1 751 9 0 9693 0 239 1 761 0 232 1 707 10 0 9727 0 284 1 716 0 276 1 669 11 0 9754 0 321 1 679
17. 9220 English PI SQC User Guide 43 Technical Support and Resources 44 Support may be provided in languages other than English in certain centers listed above based on availability of attendants If you select a local language option we will make best efforts to connect you with an available Technical Support Engineer TSE with that language skill If no local language TSE 1s available to assist you you will be routed to the first available attendant If all available TSEs are busy assisting other customers when you call you will be prompted to remain on the line to wait for the next available TSE or else leave a voicemail message If you choose to leave a message you will not lose your place in the queue Your voicemail will be treated as a regular phone call and will be directed to the first TSE who becomes available If you are calling about an ongoing case be sure to reference your case number when you call so we can connect you to the engineer currently assigned to your case If that engineer 19 not available another engineer will attempt to assist you Search Support From the OSIsoft Technical Support Web site click Search Support Quickly and easily search the OSIsoft Technical Support Web site s Support Solutions Documentation and Support Bulletins using the advanced MS SharePoint search engine Email based Technical Support techsupport osisoft com When contacting OSIsoft Technical Support by e
18. Center Line X Engineering Units Center Line N Sigma Center Line 3 Sigma Center Line 4 Sigma Center Line 5 Sigma USL LSL which uses upper and lower specification limits Absolute Eng Units which allows manual entry of maximum and minimum scale values Plot Time The plot time refers to the range of time over which the data 1s plotted and SQC calculations are assessed Several Start Time options are available Standard PI time relative expressions measuring time in multiples of hours and days from the present time Samples Before End Time allows you to specify a number of samples previous to the End Time using a field that appears below the Start Time field when this option 1s selected Select or input an integer for the number of samples The default value is 20 In Run Mode a PI SQC chart pans across values by the number of samples defined when stepping forward or backward through time Trigger Tag Change indicates that the start of each plot and product run is defined by the first data point that occurs after a specific trigger tag changes value Use the field that appears below the Start Time field when this option is selected to enter a tag name or click the ellipsis button next to the field to run a tag search In Run Mode the PI SOC Chart pans across values based on instances in time when the Trigger Tag value changes All three parts of the chart are recalculated when the value is updated
19. Distribution and Standard Deviation 8 9 E Edit a PI SQC Chart 11 Exponentially Weighted Moving Average EWMA Calculations 32 Exponentially Weighted Moving Average EWMA Chart 13 17 19 F Factors 27 36 Factors for Averages 30 31 32 36 Factors for Ranges 34 37 Factors for Standard Deviations 35 38 Filter Equation Syntax 20 Filter Equations 19 Format Tab 9 24 G General Tab 12 Glossary 41 PI SQC User Guide H Histogram 5 6 8 26 Individuals Calculations 27 Individuals Chart 12 Introduction to PI SQC 1 K Keep Bad Values 20 L Lambda 13 19 Legend 5 9 M Mathematical Symbols 27 39 Moving Average Calculations 31 Moving Average Chart 13 Moving Range Rm Chart 14 Moving Range Calculations 34 Moving Standard Deviation Calculations 35 Moving Standard Deviation Chart 13 P Pattern Tests 23 24 PI SQC Chart Components 5 PI SQC Chart Definition Dialog 11 PI SQC RT 2 3 7 14 Plot Time 9 16 22 25 Process Capability Index Cpk 9 10 26 39 R Range R Chart 14 Range Calculations 34 S Sample Period 17 19 Sample Size 17 18 Sample Tab 17 Sampling Method 17 Scale 15 Sigma 6 7 14 47 Index Sigma Calculation Data Filter 22 SQC in PI ProcessBook 2 SQC in the Process Industry 2 Standard Deviation s Chart 13 Standard Deviation Calculations 34 36 Start Time of
20. SQC Chart The cursor changes to an arrow with a small SQC label Click and drag to define the bounding rectangle for the chart within the display Once the bounding rectangle has been defined the PI SQC Chart Definition dialog appears Define the chart using the six tabs of the P SOC Chart Definition dialog page 11 Edit a PI SQC Chart To modify an existing PI SQC chart In Build mode double click anywhere in the PI SQC Chart to bring up the Chart Definition dialog page 11 In Run mode click the Item Definition button to display the Chart Definition dialog page 11 PI SQC User Guide 11 PI SQC Chart Definition Dialog General Tab The General tab resembles the PI ProcessBook Trend Definition dialog and is used to specify basic chart parameters Chart Title Enter a distinctive title for the PI SQC Chart The default is SQC Plot 1 Chart Type Select the type of control chart you want to display The name of the chart type is included in square brackets after the control chart title Each control chart presents a different view of sample data and represents different mathematical calculations page 27 based on the type of plot sample size and other factors The eight available control charts and their properties are described in the following sections Individuals Chart An Individuals chart plots raw data from PI or an ODBC dataset and utilizes single observations rather than sample groups to evaluate
21. Sample 17 19 Statistical Calculation Basis 27 Statistical Quality Control 1 T Tag Search 15 Technical Support and Resources 43 Time Stamp at 19 Transition Data Filter 23 Trend Cursors 7 8 X X Bar Average Chart 12 X Bar Calculations 29 48 os
22. Sigma test which indicates data points outside or on the upper or lower control limits Note The lower control limit in Range and Standard Deviation charts is always zero for sample sizes smaller than 7 and 6 respectively Alarm Status A control chart displays a marker for each data point If a point is in an alarm condition page 23 an alarm marker is substituted If a data point falls outside of the vertical plot range its marker 1s displayed at the top or bottom of the control chart PI SQC RT page 3 may also be used to configure alarms on the PI Server and which appear in a control chart A selected data point is in alarm status if it meets a pattern test condition A green dot indicates no pattern test conditions have been met red exclamation point m indicates a pattern test condition has been met To view alarm tag details in a control chart e Move the mouse over a data point until a magnifier cursor appears and click on the left mouse button The Alarm Details dialog appears The Alarm Details dialog lists the timestamp value and alarm status of a selected data point Note The Alarm Details dialog is disabled when trend cursors page 8 are activated If SQC alarms are configured to use the optional test status tag you can right click a marker in an alarm state to view pass fail status for all configured pattern tests If you do not use the optional test status point then only the highest
23. a Retrieval and Calculation Once a PI SQC chart is configured PI ProcessBook retrieves fresh data whenever you open the display containing the chart The time stamp of a PI tag a specified number of data points or a change in a Trigger Tag can also trigger a query Sampling is either time or event based SQC performs calculations when data is retrieved including any PI SOC RT page 3 calculations required remotely and then updates chart values The same events occur when you modify properties of the SQC calculation or chart allowing you to instantly view and analyze changes If the end time of a PI SQC chart 1s configured as the current time PI SQC refreshes the chart whenever there is a new chart tag value recorded in PI or ODBC data sources 2 os PI SQC RT PI SQC RT PI SQC RT is a separately licensed enhancement to the standard PI Alarm Subsystem PI SQC RT generates SQC Alarm tags which combine source tag values with the results of real time SQC pattern tests performed on the source tag A PI SQC alarm requires five PI tags 1n addition to the alarm tag itself These tags provide a data source for the alarm user control over the operation of the alarm tag storage of control limits PI SQC RT alarms can be plotted and visualized through the PI SQC add in See the PJ SOC RT User Guide for more information PI SQC User Guide 3 Chapter 2 PI SQC Chart Components A PI SQC chart includes several c
24. amp and point Value The Status parameter sometimes called IStat depends on the tag s Point Type in PI Plot Data Sheet The Index column identifies plotted data points followed by Timestamp and Value Following are columns for each control parameter page 21 which may vary over time The Alarm column indicates whether a data point has triggered an alarm or exceeded 1 Sigma from the center line Control Limits Sheet Specifications and calculated or specified control limits are listed 10 os Chapter 2 PI SQC Chart Definition Dialog The PI SQC Chart Definition dialog is used to define parameters and construct various types of SQC charts Note For calculated tags from PI for Open VMS Server much of the definition functionality is disabled and only Alarm and Format properties can be modified Build a Pl SQC Chart When designing a PI SQC chart keep in mind that you are simultaneously creating and structuring a query to retrieve the desired tag and SQC data If the resulting PI SQC chart is not populated with appropriate data the query has failed and the you may receive a message such as Event based query failed or Not enough data points Review your selections in the Chart Definition dialog until your PI SQC chart appears correctly To build a new PI SQC chart l p Open a ProcessBook display or an independent display file Select the PI SQC Chart tool Ed on the toolbar or choose Draw
25. apable of performing to tighter tolerances SQC 1s one approach to maintaining and improving quality SQC identifies instances of unnatural fluctuation so that causes can be assigned and corrected Control charts are employed by a wide range of industries and agencies as a means to monitor and stimulate improvements in many types of processes In many industries SQC has been applied with large economic benefits However applications of SQC methods have been limited by the daunting and often tedious tasks of calculating samples and plotting data especially when dealing with very large data sets SQC in PI ProcessBook PI SQC appears appears as an icon Reg on the PI ProcessBook toolbar To add an SQC chart object to a ProcessBook display 1 In Build mode click the SQC icon and drag in the workspace to add the object The PI SQC Chart Definition dialog appears 2 Configure chart settings and click OK 3 Return to Run Mode As a dynamic object PI SQC operates like to other dynamic symbols in PI ProcessBook displays When you select an SQC symbol and view the Details window in ProcessBook you see a tabular view of the data used to create the SQC chart For event based SQC charts based on PI points you can also create annotations for individual event values when Data is selected in the Data Source drop down list If sample filtering is used in the chart annotation 1s disabled for any rows that display the Filtered indicator Dat
26. e 2 period moving ranges are Ry Xi Xx p with n the sample size and X the individual measurements Sigma Calculation Average Standard Deviation Method The sample standard deviation o 1s calculated from individual measurements as 28 os Control Chart Calculations where the constants c and E 3 c4 are 8862 and 3 385 respectively for 3 period moving standard deviation calculations The average 3 period moving standard deviation is calculated as pDT 2 2 where 3 period moving standard deviations are g where the average for the 3 period moving subgroup k is Vict gt 2 Xi with n the sample size and X the individual measurements X X Bar Calculations Sample Average sample averages are calculated as X _ X k 1 ensi n for consecutive samples k 1 to g where X k 1 enti are individual measurements i measurement of consecutive sample k g is the number of samples subgroups n is the sample size size of each sample Center Line The Center Line 1s the grand average of the sample averages y v X g Control Limits PI SQC User Guide 29 Statistical Calculation Basis 30 The Upper and Lower Control Limits are calculated as d UCL X 3 vn g LEL X 3 vn Sigma Calculation Average Range Method The estimate of the process standard deviation sigma 1s calculated using the ave
27. e sensitive in revealing changes in process level than individual observations os General Tab Moving Average Chart A Moving Average chart calculates averages inclusive of each new individual measurement n and plots them against previous n averages Moving Average charts have the advantage of being able to detect process shifts and dampen out random errors on individual measurements and are useful when taking more than a single observation per sample is impractical In evaluating a Moving Average Chart consider only whether data 1s within 3 sigma Moving Average charts should be used in conjunction with the Moving Range chart Two disadvantages of Moving Range charts are A lag effect due to sample measurements derived from n previous measurements e Oscillatory effects that occur due to successive samples having correlated random error Sample size cannot be less than 2 in Moving Average charts Exponentially Weighted Moving Average EWMA Chart An EWMA chart is an exponentially smoothed moving average used for evaluating the stability of a process The term exponentially refers to the method that is employed to average earlier data with current data A Lambda page 19 factor is used to associate a weight with each new data point compared to the remaining data points in the designated sample size This exponential weighting 1s applied to each data point whether individual observations or subgroup avera
28. echnical data as well as a special collection of resources for system managers For these options click Knowledge Center on the Technical Support Web site e The Search feature allows you to search Support Solutions Bulletins Support Pages Known Issues Enhancements and Documentation including user manuals release notes and white papers e System Manager Resources include tools and instructions that help you manage Archive sizing backup scripts daily health checks daylight savings time configuration PI Server security PI System sizing and configuration PI trusts for Interface Nodes and more Upgrades From the OSIsoft Technical Support Web site click Contact Us gt Obtaining Upgrades You are eligible to download or order any available version of a product for which you have an active Service Reliance Program SRP formerly known as Tech Support Agreement TSA To verify or change your SRP status contact your Sales Representative or Technical Support http techsupport osisoft com for assistance PI SQC User Guide 45 Index A Alarm Status 7 Alarm Tab 7 23 26 B Build a PI SQC Chart 11 C Calculation Basis 18 Calculation Period 17 18 Chart Tag 9 14 15 20 25 Chart Type 12 Control Chart 5 Control Chart Calculations 12 27 Control Parameters Tab 6 9 10 21 Control Zones 6 21 D Data Filter Tab 22 Data Retrieval and Calculation 2 Data Sheets 5 9
29. ed in a statistical calculation to produce a single sample For moving window event based calculations a sample is produced from the raw data array R by using Ri through R n i Time based Sampling Time based sampling is used only for PI data This method organizes data into calculation packets by working with four parameters Parameter Definition Start Time page This time stamp adjusts the start time of the samples 19 Calculation Period Duration of time between the start of sampling periods page 19 Sample Period Time between samples within sample group page 19 Sample Size The count of raw data points used in the sample calculation page 18 sampling begins at the Start Time and continues with samples calculated every Calculation Period Within a Calculation Period the first n samples spaced evenly by the Sample Period are used for calculation PI SQC User Guide 17 PI SQC Chart Definition Dialog 18 For an EWMA page 13 chart a sample is produced for the first calculation period the window then the window 1s moved forward one sample period in time and the next calculation made Note An ODBC data set without a time column does not display a time stamp on the horizontal axis of a control chart Calculation Basis Calculation Basis options define the mathematical parameters used for sampling lt A Time based calculation uses samples for every Sample Period and
30. erage Range Method The estimate of the process standard deviation sigma is calculated using the average of sample ranges 3 vn A R where A is listed in the Factors for Averages page 36 table for various sample sizes n PI SQC User Guide 31 Statistical Calculation Basis Average Range The Average Range is calculated as x9 p k g where sample ranges are Ry Min X 1 with 1 1 to n in each moving sample k 1 to g Sigma Calculation Average Standard Deviation Method The estimate of the process standard deviation sigma 1s calculated using the average of standard deviations as E p As vn where A is listed in the Factors for Averages page 36 table for various sample sizes n 3 Average Standard Deviation The Average Standard Deviation is calculated as 8 mz Aci Sk g where sample standard deviations are n 1 for each sample k 1 to g with sample averages as calculated previously Exponentially Weighted Moving Average EWMA Calculations Sample Averages The averages for consecutive samples are g d i X k 1 en i TL for samples k 1 to g where X k 1 ETLE are individual measurements UT measurement of consecutive sample k g is the number of samples subgroups n is the sample Size size of each sample 32 os Control Chart Calculations If the sample size is 1 sample averag
31. eriod between the left and right axis of the chart Note For PI SQC RT Alarm Tags Data Filter information is not available os Alarm Tab Sigma Calculation Data Filter A Sigma Calculation Data Filter provides logic that determines how the system analyzes and discards data prior to performing statistical calculations Enter a Minimum data points value to specify the threshold required for a valid Sigma calculation This value is required and should depend on the sample size and the particular operating environment The default is 10 and the value must be an integer e When selected the Eliminate samples option eliminates data outside of a specified number n Sigma Generally n is set to a value between 4 and 7 The default 1s 5 and n must be an integer greater than one Transition Data Filter The Transition Data Filter allows you configure the trigger tag to specify when calculations should begin after a new trigger tag event occurs The complete configuration sentence with default values reads After trigger tag event begin calculating when 5 consecutive samples within 3 Sigma of each other are reported Four variables in the filter statement are configurable e Number of samples default is 5 e Consecutive or non consecutive samples e Range by which two data points can be separated before calculation resumes default is 3 Sigma e Variability around either the center line or between sampl
32. es to indicate the relationship data points must share default is Each Other indicating relationship between samples You may also force the system to ignore the filter statement entirely by setting the number of samples variable to 1 Use the check boxes to specify if the Transition Data Filter should apply to Sigma calculations to alarm calculations or to both Alarm Tab The Alarm tab provides tools to define pattern tests page 24 for unnatural data variation and to indicate process points that are out of statistical control PI SQC applies pattern tests and calculates an alarm state for each sample Points out of statistical control can be configured to display with specific symbols and colors in the control chart PI SQC User Guide 23 PI SQC Chart Definition Dialog 24 Note For PI SQC RT Alarm tags Alarm tab settings are read only except alarm marker type and alarm marker color The Alarm tab includes criteria and settings for seven SQC pattern tests in order of highest precedence e Outside Control A count of the number of data points outside the control limit on one side of the centerline e Outside Two Sigma A count of the number of data points outside a limit drawn two thirds of the distance between the centerline and a control limit Outside One Sigma A count of the number of data points outside a limit drawn one third of the distance between the centerline and a control limit
33. es x k are individual measurements Transformation of individual observations The consecutive sample averages are transformed into a series of weighted averages Lk AX 1 m Zx 41 fork 1tog with 4 Xi where A is the weighting factor Center Line The Center Line 1s the average of the transformed data and calculated as g Control Limits z The equations needed for calculating the upper and lower Control Limits are similar to as those used for the Individuals charts except that they take into account the weighting factor Sigma Calculation Average Moving Range Method UCL Z LR LCL Z LR where Because EWMA involves two period calculations 4 1 128 E 2 66 and for various values of A L is calculated as A 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 L 0 610 0 887 1 117 1 330 1 536 1 741 1 952 2 172 2 175 Average Range PI SQC User Guide 33 Statistical Calculation Basis The Average Range is calculated as R o Ry g where sample ranges are Min X i 1 with 1 1 to n in each moving sample k 1 to g Range Calculations Sample Ranges The ranges of individual samples are Max X x 1 n i Min X 2 for i 1 to nin each sample k 1 to g Center Line The Center Line is calculated from sample ranges as Control Limits Upper Control Limit UCL D R Lower Contr
34. ges and has the effect of reducing the contribution of older data through progressively smaller weights The EWMA chart can be used in place of a Moving Average chart to place more emphasis on current data values Placing more emphasis on current data makes the EWMA chart more sensitive to small process trends and less sensitive to sudden process spikes Note The Lambda weighting factor is typically between 0 2 and 0 5 with a default value of 0 5 Standard Deviation s Chart A Standard Deviation chart calculates and plots a sample standard deviation for each sample subgroup A Standard Deviation chart serves the same purpose as the Range chart but should be used when the sample size 19 greater than 8 Note Sample size must be greater than 1 and preferably equal to or greater than 10 the default sample size PI SQC User Guide 13 PI SQC Chart Definition Dialog 14 Moving Standard Deviation Chart A Moving Standard Deviation chart utilizes the latest n observations to judge the stability of the process The current sample standard deviation 1s calculated by replacing the oldest value in the sample with the current value Moving Standard Deviation charts are used instead of Moving Range charts when the sample subgroup size 1s greater than 8 As with Moving Average charts the effect of combining successive data points causes correlated contributions of random error and results in an oscillatory effect when plotted
35. hich may either be process specifications or control limits e Process specifications are fixed not to exceed limits applied to a product and based on market safety or engineering limits Control limits are calculated from measurements sampled during a controlled process operation or from the actual data of the current process Note For PI SQC RT Alarm Tags control limits are read only Refer to the P SQC RT User Guide for information on remote calculation of control limits in Pl All PI SQC control limits used to define control chart zones page 6 are accessible through the Control Parameters tab You can select constants PI tags or ODBC data sets to define limits for each parameter Enter constant values directly in Control Limit fields PI SQC User Guide 21 PI SQC Chart Definition Dialog 22 If constants are used then the control limits do not change unless the user returns to the Control Parameters tab and manually changes them Click Tag Search to use the Tag Search dialog to find and select PI tags Corresponding tag values are used as control limits Click the arrow button adjacent to Tag Search and select PI Calculation or ODBC to display a corresponding Data dialog Then choose a dataset that contains the desired control limit values If an ODBC data set 1s used for limits you must specify Start and End Times under Plot Time page 16 The PI ProcessBook User Guide includes information on how to
36. ion Limits pen color and line style for USL LSL Histogram Bars color of histogram bars Histogram Distribution pen color and line style for the curve plotted on the Histogram Marker shape and color of the data markers that are not in alarm A value of None causes all markers to disappear Histogram Specs pen color and line style for specification lines on the histogram Histogram Mean pen color and line style for the mean on the histogram Control Limits pen color and line style for the UCL and LCL on the control chart Chart Options The check boxes in the Chart Options group designate whether or not to display the following items in the PI SQC chart Chart Title Histogram Legend Control Chart PI SQC User Guide 25 PI SQC Chart Definition Dialog 26 e Control Limit Scale Plot Limits Stepwise Legend The Legend group check boxes designate whether or not to display the following items in the chart legend e Chart Tag the name of the chart tag page 14 e Trigger Tag the name of the trigger tag page 16 if used e Eng Units the engineering units associated with the chart tag e USL LSL upper and lower specification limits e Value the current value of the chart tag e Sigma the sigma value of control limits the Cpk see Process Capability Index Cpk on page 40 based on the current plot Zone Colors Zone color samples appear at the upper right of the tab below shape and color sa
37. lack one or more of the three characteristics of a natural pattern e OSI Appendix C Technical Support and Resources You can read complete information about technical support options and access all of the following resources at the OSIsoft Technical Support Web site http techsupport osisoft com http techsupport osisoft com Before You Call or Write for Help When you contact OSIsoft Technical Support please provide e Product name version and or build numbers e Computer platform CPU type operating system and version number The time that the difficulty started e The message log s at that time Help Desk and Telephone Support You can contact OSIsoft Technical Support 24 hours a day Use the numbers in the table below to find the most appropriate number for your area Dialing any of these numbers will route your call into our global support queue to be answered by engineers stationed around the world Office Location Access Number Local Language Options San Leandro CA USA 1 510 297 5828 English Philadelphia PA USA 1 215 606 0705 English Johnson City TN USA 1 423 610 3800 English Montreal QC Canada 1 514 493 0663 English French Sao Paulo Brazil 55 11 3053 5040 English Portuguese Altenstadt Germany 49 6047 9890 English German Manama Bahrain 973 1758 4429 English Arabic Singapore 65 6391 1811 English Mandarin 86 021 2327 8686 Mandarin Perth WA Australia 61 8 9282
38. lculated as sk n 1 for each sample k 1 to g where AL i 1 are individual measurement 1 1 ton g is the number of samples subgroups n is the sample size size of each sample sample averages X k are calculated as y S zi k isi E n for consecutive samples k 1 to g Center Line The Center Line is the Average Standard Deviation which 1s calculated from sample standard deviations as k 1 k g Control Limits j See Standard Deviation Calculations page 35 for upper and lower control limits Factors The following sections detail factors used in control chart calculations In all tables n is the sample size Factors for Averages Observations in Sample n n A2 A3 2 1 880 2 659 3 1 023 1 954 4 0 729 1 628 36 os n A2 A3 5 0 577 1 427 6 0 483 1 287 7 0 419 1 182 8 0 373 1 099 9 0 337 1 032 10 0 308 0 975 11 0 285 0 927 12 0 266 0 886 13 0 249 0 850 14 0 235 0 817 15 0 223 0 789 16 0 212 0 763 17 0 203 0 739 18 0 194 0 718 19 0 187 0 698 20 0 180 0 680 21 0 173 0 663 22 0 167 0 647 23 0 162 0 633 24 0 157 0 619 25 0 153 0 606 Factors for Ranges Observations in Sample n n d2 D1 D2 D3 DA 2 1 128 0 3 686 0 3 267 3 1 693 0 4 358 0 2 574 4 2 059 U 4 698 0 2 282 9 2 326 0 4 918 0 2 114 6 2
39. ld e Click Tag Search to locate PI tags and append the selected result to the equation in the Filter Equation box For multiple selected items only the first is added PI tags used in a filter equation must exist on the same PI server as the Chart Tag page 14 e Click Check Syntax to send the equation to the PI server to validate syntax page 20 A dialog box indicates any errors in the equation syntax The contents of the Filter Equation field are also validated whenever you exit the Sample tab e Clear the Keep Bad Values checkbox to omit bad data values page 21 from samples and from control limit calculations Filter Equation Syntax Filter equations use PI Performance Equation syntax and may include operands e arithmetic operators e built in functions e user defined formulas e if then else structures PI tag values Filters may cause large time gaps in data To avoid gaps you may choose to keep bad values page 21 Note When defining a filter equation use PI tags with the step attribute set to TRUE Interpolation of tag values in the filter expression might otherwise provide a misleading visualization os Control Parameters Tab Keep Bad Values The number of plot markers in a time series is the number of good values divided by sample size If they do not divide evenly the remainder is ignored For example given a sample size of 10 and 99 good values a plot will have 9 points
40. mail it is helpful to send the following information e Description of issue Short description of issue symptoms informational or error messages history of issue e Message logs See documentation for your PI System for information on obtaining message logs pertinent to the situation Online Technical Support From the OSIsoft Technical Support Web site click Contact us gt My Support gt My Calls Using OSIsoft s Online Technical Support you can Enter anew call directly into OSIsoft s database monitored 24 hours a day e View or edit existing OSIsoft calls that you entered View any of the calls entered by your organization or site if enabled e See your licensed software and dates of your Service Reliance Program agreements os Remote Access From the OSIsoft Technical Support Web site click Contact Us gt Remote Support Options OSIsoft Support Engineers may remotely access your server in order to provide hands on troubleshooting and assistance See the Remote Access page for details on the various methods you can use On site service From the OSIsoft Technical Support Web site click Contact Us gt On site Field Service Visit OSIsoft provides on site service for a fee Visit our On site Field Service Visit page for more information Knowledge Center From the OSIsoft Technical Support Web site click Knowledge Center The Knowledge Center provides a searchable library of documentation and t
41. mples of data markers used for points in alarm as defined on the Alarm tab page 23 Zone colors are set with other Chart Element controls but you can make final adjustments to alarm markers using the data marker buttons Histogram Bars Defines the number of bars to be included in the histogram page 8 Save as Default To retain your format settings as the default formatting for future charts click the Save as Default button below the Chart Options The settings are stored in the Procbook ini file until they are saved again os Appendix A Statistical Calculation Basis These general conventions apply to statistical calculations For PI tags data precision is set to a tag s display precision property in PI For ODBC data sets precision is set to 5 significant figures Ifa center line is not specified for a chart the grand average of data 15 used instead Either a constant or PI tag may be specified as a control limit If control limits are not specified theoretical control limits are calculated from available data using the ANSI ASQC standard A1 1987 e Control limits are calculated separately to allow control limit bands of different heights The statistical analysis methods used by PI SQC are based on concepts and equations from Essentials of SPC in the Process Industries by James M Pruett and Helmut Schneider Second printing February 1996 and the ANSI ASQC standard A1 1987 Control Chart Calculatio
42. ndividual cases and events are unpredictable Y ou cannot predict exactly how long you will live individual points are not predictable Groups within a constant system of causes tend to be predictable Actuaries can use population data to predict life expectancy for insurance companies a series of points from a constant process tend to follow a pattern A series of measurements of a process parameter generally will not be identical to each other However you can use statistical methods to establish baselines usually called control limits using data that represent good operating conditions Statistically derived control limits outline the potential behavior of a process and permit differentiation between random fluctuations of data and true process shifts A control limit can be defined by a specific number center line 3 standard deviations or in terms of pattern tests 4 successive points greater than 1 5 standard deviations from the center line on the same side of the center line The primary method of process evaluation is a control chart a graphic comparison of process data versus calculated control limits You can use a control chart to detect process variations that have definable causes by highlighting instances where process data runs outside the control limits PI SQC User Guide 1 Introduction to PI SQC SQC in the Process Industry Product specifications typically are market based and an actual process is c
43. ns The following sections detail the calculations used in plotting PI SQC control charts Statistical formulas presented in this section refer to constants page 36 and symbols page 39 described in corresponding topics Individuals Calculations Sample Average The sample average value is calculated using the equation Y Lii Xi Tl where X are individual measurements n is the number of measurements Control Limits PI SQC User Guide 2 Statistical Calculation Basis The three sigma control limits for Individuals charts are UCL X 30 LCL X 3o where UCL is the Upper Control Limit LCL is the Lower Control Limit X is the sample average as calculated previously o is the process standard deviation also known as sigma The process standard deviation is estimated using one of two approaches Average Range Estimation method Calculate an estimate of the process standard deviation based on the average of moving ranges e Average Standard Deviation Estimation method Calculate an estimate of the process standard deviation based on the average of moving standard deviations Sigma Calculation Average Range Method Using this method the process sigma is derived as follows R 203 d where the constants d and E 3 42 are 1 128 and 2 660 respectively for two period moving range calculations The two period average moving range is calculated as n ln 2 1 Ry 1 wher
44. ol Limit LCL D R D The constants P3 and P are listed in the Factors for Ranges page 37 table for various sample sizes n Moving Range Calculations Sample Ranges The ranges of individual samples are Max Xx i 1 with 1 1 to n in each moving sample k 1 to g Center Line The Center Line is calculated from sample ranges as g J 2 Rx g R 34 os Control Chart Calculations Control Limits See Range Calculations page 34 for upper and lower control limits Standard Deviation Calculations Sample Standard Deviations The standard deviations of individual samples are calculated as 1 in each sample k 1 to g where ALK enr are individual measurements for 1 1 ton g 1s the number of samples subgroups n is the sample size size of each sample sample averages X k are calculated as NN TL for consecutive samples k 1 to g Center Line The Center Line is the Average Standard Deviation which 1s calculated from sample standard deviations as _ 4k 15k g Control Limits Upper Control Limit Vol a Lower Control Limit LCL 3 and are listed in the Factors for Standard Deviations page 38 table for various sample sizes n PI SQC User Guide 35 Statistical Calculation Basis Moving Standard Deviation Calculations Sample Standard Deviations The standard deviations of individual samples are ca
45. omponents A control chart on the bottom A histogram page 8 at top left A legend page 9 at top right Four sheets of data and statistics page 9 accessible when you click the chart title You may configure the control chart the histogram and the legend in a variety of ways and any component may be omitted Control Chart A control chart shows a curve of data markers representing sample values plotted across time PI SQC User Guide The left axis displays evenly spaced labels in the engineering units of the source tag The right axis displays control limits The horizontal axis displays evenly spaced data markers PI SQC Chart Components 122 734 e 19 158 E 115 578 P N H12 108 422 104 844 101 266 10 22 2007 7 41 05 10 22 2007 9 11 06 AM x The left and right timestamps on the horizontal axis represent the first and last plot values Each plot point on a control chart is equidistant from preceding and succeeding points regardless of the time span between them If a sample value plots outside of the control limits an alarm condition occurs and data markers appear in different colors and shapes Data markers are not displayed if they overlap only alarm markers appear A green triangle at lower right indicates the PI SQC chart is being updated in real time Control Zones Control charts include colored horizontal control zones and a target or center line The colors match
46. ot configure specification limits page 21 the USL LSL is not displayed Calculated items such as Cpk or Sigma that do not have a current value will display N A if they are selected to show on the legend To change the appearance of the legend use the Format tab page 25 of the PI SQC Chart Definition dialog Data Sheets Four data sheets are available for each PI SQC Chart To see the data sheets 1 Goto Run Mode and double click the title of the chart The Statistical Quality Control Details dialog appears 2 Use the Options menu to select any data sheet PI SQC User Guide 9 PI SQC Chart Components Statistical data are provided in double precision numbers Click Save Data to File to save all four pages as a text file Statistics Sheet Eight statistical values are calculated for control chart plot data each time a PI SQC chart updates Statistic Description Mean Average value of plot data Median Median value of plot data Mode Most frequently occurring value of plot data STDEV Standard deviation of plot data Cpk Calculated Process Capability Index Cpk see Process Capability Index Cpk on page 40 based on specification limits for plot data Max Maximum value of plot data Min Minimum value of plot data Sigma Sigma of the samples as used in upper and lower control limits Raw Data Sheet The Index column identifies each raw data point plotted for the chart tag followed by Timest
47. precedence pattern test failure 1s displayed For information on how to configure PI SQC RT Alarm tags refer to the PI SOC RT User Manual PI SQC User Guide 7 PI SQC Chart Components Trend Cursors In PI ProcessBook Run Mode trend cursors can be added from the left edge of a control chart to read sample values across the chart Add a second cursor to compare values across a range To add trend cursors e Click the Trend Cursor button on the toolbar to add a cursor to the right edge of the chart When released a trend cursor remains in one spot Click and drag a cursor to move it to a different spot 223505 13 003 15 502 12 108 488 104 337 101485 222310 2272007 8 05 06 AM 10 22 2007 7 45 06 Histogram The histogram component of a PI SQC chart 1s a bar graph of the frequency distribution of plot data 101 268 122734 The horizontal scale spans the Center Line minus five Sigma to the Center Line plus five Sigma The scale is marked at 3 Sigma and 3 Sigma from the center line The resulting zones are colored using the same colors in the control chart The Mean Upper and Lower Specification Limits are also indicated in the histogram The number of bars is configurable and from 1 to 100 bars may be displayed The vertical scale runs from zero to the maximum number of samples per bar A normal distribution curve page 9 1s drawn in the histogram based on the sample mean and the s
48. rage of sample ranges a R 4n nd A R where A is listed in the Factors for Averages page 36 table for various sample sizes n Average Range The Average Range calculated as x R p k l g where sample ranges are Ry Max X k 1 enti Min Xti 12i with 1 1 to n in each consecutive sample k 1 to g Sigma Calculation Average Standard Deviation Method The estimate of the process standard deviation sigma 1s calculated using the average of standard deviation as 0 Vn where A3 is listed in the Factors for Averages page 36 table for various sample sizes n A S Average Standard Deviation The Average Standard Deviation is calculated as Lii Sk g where sample standard deviations are os Control Chart Calculations Sk T n 1 for each sample k 1 to g with sample averages as calculated previously Moving Average Calculations sample Averages sample averages are calculated as a n for each moving sample k 1 to g where are individual measurements ith measurement of moving sample k g is the number of samples subgroups n is the sample size size of each sample Center Line The Center Line 1s the average of sample averages ag E 24 Xk g Control Limits y The Upper and Lower Control Limits are calculated as shown below UCL lt X 3 vn g LCL X 3 vn Sigma Calculation Av
49. rocessPoint Sigmafine Analysis Framework IT Monitor MCN Health Monitor PI System PI ActiveView PI ACE PI AlarmView PI BatchVievv PI Data Services PI Manual Logger PI ProfileView PI Web Parts ProTRAQ RLINK RtAnalytics RtBaseline RtPortal RtPM RtReports and RtWebParts are all trademarks of OSIsoft Inc All other trademarks or trade names used herein are the property of their respective owners U S GOVERNMENT RIGHTS Use duplication or disclosure by the U S Government is subject to restrictions set forth in the OSIsoft Inc license agreement and as provided in DFARS 227 7202 DFARS 252 227 7013 FAR 12 212 FAR 52 227 as applicable OSIsoft Inc Published 9 2 2009 Table of Contents Chapter 1 Introduction to PI SQC uuu sees eee eee 1 StaliStiGall evi Eeeni 1 S0C in ne ue AeiiM 2 CEFE See oo ANT TT a een ee 2 Fre UT 3 Chapter 2 PI SQC Chart Components 5 522562 2 ee eee eee 5 CONTO OA 5 SO n s b S 8 Mole 1070751757 ee 9 D 12 TTT 9 Chapter 3 PI SQC Chart Definition 1 2 52225 ennenen neee 11 Bulda PUL a e 11 eG Bee ua 11 err ME MMNRDEc 12 ecl b ii b 17 Control Parameters TAL aa Ra aaa ENN ENAR 21 Buceo i 22 P mnEE o T
50. s data being plotted Tag names or data sets that define PI SQC limits or parameters are defined on another tab You can enter this tag manually or select the tag from PI through a Tag Search Tag names returned from a Tag Search appear with an appended server name server tagname in the Chart Tag box os General Tab Note The Chart Tag field must be filled before you can switch to another tab The Chart Tag field is also used for PI SOC RT page 3 Alarm tags which are configured at the PI server level where the point class is SOC Alarm Click Alarm Tag Attributes to view the attributes of an Alarm tag The query that retrieves Chart Tag page 14 values also retrieves the point attributes shown below Point Attribute Name Usage Descriptor Displayed in the Legend Point Type Used in calculations not displayed Zero May be used as lower plot limit Span Defines data range as vertical scale Starting digital state Digital zero Number of digital states Number of possible digital states minus one Engineering units Displayed in the Legend Tag Search A PI tag search allows you to search all connected PI Servers for tags meeting a given a set of criteria such as one ore more tag attribute values e Click Tag Search to search a PI Server for tags using the Tag Search dialog You can select tags from a Search Results list The first tag selected 1s added to the Chart Tag field with
51. tandard deviation of the sample 8 osi Legend To specify the number of bars or to change the appearance of the histogram use the Format tab page 25 of the PI SQC Chart Definition dialog Distribution and Standard Deviation If you specify control limits page 21 the Sigma that 1s used in the normal distribution curve is calculated by 1 3 the distance between the center line and each control limit You may specify uneven control limits that is the distance of the upper control limit from the center line may be different from the distance of the lower limit from the center line In such cases the normal distribution curve reflects Sigma values obtained by Sigma Upper Control Limit Center Line 3 Sigmaj Center Line Lower Control Limit 3 Legend The legend appears at top right in a PI SQC chart and may display the following information Item Description Chart tag The name of the tag providing raw data to the chart Trigger Tag The name of the tag used to trigger a new sampling page 16 Value The current value of the Chart Tag page 14 Engineering Units Engineering units for the Chart Tag USL LSL Upper and lower specification limits Sigma Sigma as used in calculating upper and lower control limits Cpk Process Capability Index page 40 a calculation based on specification limits Items that do not have a current value are not displayed For example if a user does n
52. the control limits on the right axis and in the histogram page 8 Zones are determined by control limits page 21 for process evaluation which may be established in several ways e Specified by a user Calculated from actual point values e Calculated remotely by a PI SQC RT Alarm tag The zones represent standard SQC classifications Center Zone The area between 1 and 1 Sigma of Center Line One Sigma Zone The area between 2 and 1 Sigma of Center Line and the area between 1 and 2 Sigma of Center Line Two Sigma Zone The area between 3 and 2 Sigma of Center Line and the area between 2 and 3 Sigma of Center Line Outside 3 Sigma Zone The areas outside control limits Note For Range Moving Range Standard Deviation and Moving Standard Deviation charts the upper and lower zones may be different in size because the Sigma page 7 values may differ Control Chart Sigma Since the target for Range Moving Range Standard Deviation and Moving Standard Deviation charts is not always centered on the chart upper and lower Sigmas may differ Sigma should be interpreted in Range and Standard Deviation charts to mean 1 3 the distance between the center line and the upper or lower control limit For instance 2 Sigma refers to two thirds of the distance between the center line and the upper control limit The most applicable pattern test for Range and Standard Deviation charts is the 3
53. ty Index Cpk or process potential refers to the ability of a process to produce products that meet customer specifications The larger the Cpk number the more likely a process 1s capable of producing product within specifications Cpk is calculated as follows Upper and lower control limits are assumed to be different One sided capability indexes are determined by Cpu USL u 36 Cpl u LSL 36 where u is process mean O is process standard deviation Then the Cpk capability index 1s determined by Cpk Min Cpu Cpl If in the rare case the standard deviation is zero the process capability index cannot be calculated and the Cpk 15 set to N A os Appendix B Glossary Fluctuating Measurements All processes have parameters that can be measured Repeated measurement of a process s parameters will show fluctuations in the parameter s value Fluctuations can be classified as natural or unnatural Natural Pattern A pattern of measurements of a process parameter exhibiting natural variability due to minute fluctuations in raw materials equipment measurement precision etc Obviously natural fluctuations result in natural patterns A natural pattern will always have the all of the following characteristics 1 Most of the points are near the Center Line 2 A few of the points spread out and approach the Control Limits None or at least only a very rare point exceeds the Control Limits Process A
Download Pdf Manuals
Related Search